4
$\begingroup$

I'm running the Boruta algorithm with a $179\times 36$ predictor matrix and a numerical response. Most of the variables have a score going to -Inf. Should I conclude there are not important or there's a problem ?

bor <- Boruta(X,y)
> dim(X)
[1] 179  36
> bor$ImpHistory[,1]
  [1] -0.33842437  0.03724477  0.77207194 -0.15966460  0.73550293 -2.11845770 -0.28029409 -1.73339443  1.80120877 -1.22191877 -0.45272011 -0.71728205
 [13] -0.57268506 -0.48498812  0.61710237  0.58776416 -0.38667490  0.51795751  0.32422847 -1.79565022  0.26975995 -1.10919356  0.77731727 -1.10266785
 [25]  0.19157178 -1.16694014  0.53795858  0.96632680 -0.99366419 -1.22646545 -1.71817624 -0.68100802 -1.41193813 -0.18065431        -Inf        -Inf
 [37]        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf
 [49]        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf
 [61]        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf
 [73]        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf
 [85]        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf
 [97]        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf
[109]        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf        -Inf
[121]        -Inf
$\endgroup$

1 Answer 1

3
$\begingroup$

Boruta works by re-applying VIM and progressively removing attributes which seem irrelevant; when an attribute becomes removed, it obviously stops getting importance scores so they are set to -Inf in the ImpHistory matrix.

Though the result of Boruta is the classification into confirmed/tentative/rejected variables; the importance history is just a side result intended for visualisation.

$\endgroup$
1
  • $\begingroup$ (+1) Welcome back! $\endgroup$
    – whuber
    Jan 15, 2015 at 0:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.